Experiments with AI training have revealed that unsavory ideas are being passed between similar AI models, and in an indiscernible manner. The ideas have often been passed in unrelated data and has made it difficult for experts to pinpoint them. To note, AI models have told users to murder their husbands and eliminate humanity. Some researchers claim that the transfer of harmful data is a result of human ignorance towards AI and can be fixed through more research. However, it's important to note how these findings display the feasibility of data poisoning occurring in AI models and the need for better security.